1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
|
/**
* Tests that a donor resumes coordinating a migration if it fails over after creating the
* migration coordinator document but before deleting it.
*
* @tags: [requires_fcv_44]
*/
// This test induces failovers on shards.
TestData.skipCheckingUUIDsConsistentAcrossCluster = true;
(function() {
'use strict';
load("jstests/libs/fail_point_util.js");
load('jstests/libs/parallel_shell_helpers.js');
function getNewNs(dbName) {
if (typeof getNewNs.counter == 'undefined') {
getNewNs.counter = 0;
}
getNewNs.counter++;
const collName = "ns" + getNewNs.counter;
return [collName, dbName + "." + collName];
}
const dbName = "test";
var st = new ShardingTest({shards: 2, rs: {nodes: 2}});
assert.commandWorked(st.s.adminCommand({enableSharding: dbName}));
assert.commandWorked(st.s.adminCommand({movePrimary: dbName, to: st.shard0.shardName}));
function runMoveChunkMakeDonorStepDownAfterFailpoint(
failpointName, shouldMakeMigrationFailToCommitOnConfig, expectAbortDecisionWithCode) {
const [collName, ns] = getNewNs(dbName);
jsTest.log("Running migration, making donor step down after failpoint " + failpointName +
"; shouldMakeMigrationFailToCommitOnConfig is " +
shouldMakeMigrationFailToCommitOnConfig + "; expectAbortDecisionWithCode is " +
expectAbortDecisionWithCode + "; ns is " + ns);
// Wait for mongos to see a primary node on the primary shard, because mongos does not retry
// writes on NotMaster errors, and we are about to insert docs through mongos.
awaitRSClientHosts(st.s, st.rs0.getPrimary(), {ok: true, ismaster: true});
// Insert some docs into the collection so that the migration leaves orphans on either the
// donor or recipient, depending on the decision.
const numDocs = 1000;
var bulk = st.s.getDB(dbName).getCollection(collName).initializeUnorderedBulkOp();
for (var i = 0; i < numDocs; i++) {
bulk.insert({_id: i});
}
assert.commandWorked(bulk.execute());
// Shard the collection.
assert.commandWorked(st.s.adminCommand({shardCollection: ns, key: {_id: 1}}));
if (shouldMakeMigrationFailToCommitOnConfig) {
// Turn on a failpoint to make the migration commit fail on the config server.
assert.commandWorked(st.configRS.getPrimary().adminCommand(
{configureFailPoint: "migrationCommitVersionError", mode: "alwaysOn"}));
}
jsTest.log("Run the moveChunk asynchronously and wait for " + failpointName + " to be hit.");
let failpointHandle = configureFailPoint(st.rs0.getPrimary(), failpointName);
const awaitResult = startParallelShell(
funWithArgs(function(ns, toShardName, expectAbortDecisionWithCode) {
if (expectAbortDecisionWithCode) {
assert.commandFailedWithCode(
db.adminCommand({moveChunk: ns, find: {_id: 0}, to: toShardName}),
expectAbortDecisionWithCode);
} else {
assert.commandWorked(
db.adminCommand({moveChunk: ns, find: {_id: 0}, to: toShardName}));
}
}, ns, st.shard1.shardName, expectAbortDecisionWithCode), st.s.port);
failpointHandle.wait();
jsTest.log("Make the donor primary step down.");
assert.commandWorked(
st.rs0.getPrimary().adminCommand({replSetStepDown: 10 /* stepDownSecs */, force: true}));
failpointHandle.off();
jsTest.log("Allow the moveChunk to finish.");
awaitResult();
if (expectAbortDecisionWithCode) {
jsTest.log("Expect abort decision, so wait for recipient to clean up the orphans.");
assert.soon(() => {
return 0 === st.rs1.getPrimary().getDB(dbName).getCollection(collName).count();
});
} else {
jsTest.log("Expect commit decision, so wait for donor to clean up the orphans.");
assert.soon(() => {
return 0 === st.rs0.getPrimary().getDB(dbName).getCollection(collName).count();
});
}
// The data should still be present on the shard that owns the chunk.
assert.eq(numDocs, st.s.getDB(dbName).getCollection(collName).count());
jsTest.log("Wait for the donor to delete the migration coordinator doc");
assert.soon(() => {
return 0 ===
st.rs0.getPrimary().getDB("config").getCollection("migrationCoordinators").count();
});
if (shouldMakeMigrationFailToCommitOnConfig) {
// Turn off the failpoint on the config server before returning.
assert.commandWorked(st.configRS.getPrimary().adminCommand(
{configureFailPoint: "migrationCommitVersionError", mode: "off"}));
}
}
//
// Decision is commit
//
runMoveChunkMakeDonorStepDownAfterFailpoint("hangBeforeMakingCommitDecisionDurable",
false /* shouldMakeMigrationFailToCommitOnConfig */);
runMoveChunkMakeDonorStepDownAfterFailpoint("hangBeforeSendingCommitDecision",
false /* shouldMakeMigrationFailToCommitOnConfig */);
runMoveChunkMakeDonorStepDownAfterFailpoint("hangBeforeForgettingMigrationAfterCommitDecision",
false /* shouldMakeMigrationFailToCommitOnConfig */);
//
// Decision is abort
//
runMoveChunkMakeDonorStepDownAfterFailpoint("moveChunkHangAtStep3",
false /* shouldMakeMigrationFailToCommitOnConfig */,
ErrorCodes.OperationFailed);
runMoveChunkMakeDonorStepDownAfterFailpoint("moveChunkHangAtStep4",
false /* shouldMakeMigrationFailToCommitOnConfig */,
ErrorCodes.OperationFailed);
runMoveChunkMakeDonorStepDownAfterFailpoint("moveChunkHangAtStep5",
false /* shouldMakeMigrationFailToCommitOnConfig */,
ErrorCodes.OperationFailed);
runMoveChunkMakeDonorStepDownAfterFailpoint("hangBeforeMakingAbortDecisionDurable",
true /* shouldMakeMigrationFailToCommitOnConfig */,
ErrorCodes.StaleEpoch);
runMoveChunkMakeDonorStepDownAfterFailpoint("hangBeforeSendingAbortDecision",
true /* shouldMakeMigrationFailToCommitOnConfig */,
ErrorCodes.StaleEpoch);
runMoveChunkMakeDonorStepDownAfterFailpoint("hangBeforeForgettingMigrationAfterAbortDecision",
true /* shouldMakeMigrationFailToCommitOnConfig */,
ErrorCodes.StaleEpoch);
st.stop();
})();
|
